AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Towards Atoms of Large Language Models

Chenhui Hu, Pengfei Cao, Yubo Chen, Kang Liu, Jun Zhao

ICML 2026 · Large language models › Language, safety and interpretability · approach: Mechanistic analysis of attention and representations · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use atomic inner product instead of Euclidean inner productpairwise representation angle analysis · 128 WikiData entities, all layers of GPT-J, Pythia, Llama, Gemmacentroid of pairwise angles: deviates from 90 deg -> restored to 90 degImproved
use TSAE atoms instead of SAE featuresrepresentational units · CounterFact, Gemma2/Llama3.1faithfulness R2: 48.8% -> 99.90%Improved
use TSAE atoms instead of SAE featuresrepresentational units · CounterFact, Gemma2/Llama3.1stability q*: 68.2% -> 99.85%Improved
use TSAE atoms instead of neuronsrepresentational units · CounterFact, Gemma2/Llama3.1stability q*: 0.5% -> 99.85%Improved
raise TSAE capacity above a critical threshold for a given data scale (ablation)TSAE on Gemma2-2B layer 1 · WikiData activations (1.9B samples)faithfulness R2Improved
choose SAE capacity heuristically regardless of data scalestandard SAE training · Gemma2-2Bfaithfulness R2Hurt
vary training hyperparameters of TSAE (ablation)TSAE atoms · Gemma2/Llama3.1faithfulness and stabilityNo clear effect
use atoms instead of SAE featuresrepresentational units · Gemma2-2B/9B, Llama3.1-8Bmonosemanticity score (LLM judge)Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy